Based on the project of Tongding Expressway in Gansu Province,the correlation between loess collapsibility grade and single physical index is studied in this paper,and the physical index with strong cor-relation are determined by combining the results of a typical loess area of a road section and the field loess physical experiment.The BP neural network quantitative prediction model of collapsibility coefficient of lo-ess applicable to the region was constructed on this basis.The results show that(1)the correlation between collapsibility coefficient of loess and natural water content,saturation,porosity,pore ratio,dry density and compression coefficient is strong,and the correlation with plasticity index and compression modulus is weak.(2)When the pore ratio e＜0.825,compression coefficient a＜0.244 and natural water content w＞18.2％,it can be judged that this type of soil does not have collapsibility.(3)The BP neural network can predict the collapsibility accurately and reliably,and its accuracy can meet the practical engineering ap-plications.The research results can be used as a theoretical basis for research and field discrimination in terms of accurate evaluation of loess collapsibility grade in this area and similar projects,and has strong theoretical and practical application value.
关键词
黄土湿陷/物性指标/BP神经网络/预测模型
Key words
Loess collapsibility/Physical properties index/BP neural net work/Prediction model
维普
万方数据